Method for training semantic representation model, device and storage medium

An electronic device, comprising:

• at least one processor;
• and
• a memory connected with the at least one processor communicatively;
• wherein the memory stores instructions executable by the at least one processor to cause the at least one processor to perform a method for training a semantic representation model, which comprises:
• obtaining an anchor sample which is a sentence including multiple tokens;
• obtaining a positive sample and a negative sample based on syntactic information of the sentence, comprising:
• constructing a syntactic tree with the multiple tokens each forming a node of the syntactic tree based on the syntactic information of the sentence;
• acquiring a first text including a first number of consecutive tokens forming a subtree in the syntactic tree, and taking the first text as the positive sample;
• and
• acquiring a second text including a second number of consecutive tokens including at least a token in the subtree and a token outside the subtree, and taking the second text as the negative sample, wherein the second number is equal to the first number;
• inputting the anchor sample, the positive sample and the negative sample into the semantic representation model, respectively, and outputting by the semantic representation model an anchor-sample semantic representation, a positive-sample semantic representation and a negative-sample semantic representation, respectively;
• constructing a contrast loss function based on the anchor-sample semantic representation, the positive-sample semantic representation, and the negative-sample semantic representation;
• and
• training the semantic representation model based on the contrast loss function, so that the trained semantic representation model has a capability of outputting a semantic representation of an inputted sentence including implicitly syntactic information of the inputted sentence.

A non-transitory computer readable storage medium storing computer instructions for causing a computer to perform a method for training a semantic representation model, which comprises:

• obtaining an anchor sample which is a sentence including multiple tokens;
• obtaining a positive sample and a negative sample based on syntactic information of the sentence, comprising:
• constructing a syntactic tree with the multiple tokens each forming a node of the syntactic tree based on the syntactic information of the sentence;
• acquiring a first text including a first number of consecutive from tokens forming a subtree in the syntactic tree, and taking the first text as the positive sample;
• and
• acquiring a second text including a second number of consecutive tokens including at least a token in the subtree and a token outside the subtree, and taking the second text as the negative sample, wherein the second number is equal to the first number;
• inputting the anchor sample, the positive sample and the negative sample into the semantic representation model, respectively, and outputting by the semantic representation model an anchor-sample semantic representation, a positive-sample semantic representation and a negative-sample semantic representation, respectively;
• constructing a contrast loss function based on the anchor-sample semantic representation, the positive-sample semantic representation, and the negative-sample semantic representation;
• and
• training the semantic representation model based on the contrast loss function, so that the trained semantic representation model has a capability of outputting a semantic representation of an inputted sentence including implicitly syntactic information of the inputted sentence.